1. Federal Department of Home Affairs FDHA Federal Office of Meteorology and Climatology MeteoSwiss Component testing of the COSMO model’s turbulent diffusion scheme Balázs Szintai MeteoSwiss, Zürich, Switzerland Advisors: Mathias Rotach, Pirmin Kaufmann Oliver Fuhrer, Matteo Buzzi MeteoSwiss COSMO General Meeting – Plenary Session 17 September 2008, Krakow, Poland

2. 2 Component testing of the COSMO model’s turbulent diffusion scheme Balázs Szintai (balazs.szintai@meteoswiss.ch) Outline PBL height determination Ideal convective case Real case study (LITFASS) Further plans

3. 3 Component testing of the COSMO model’s turbulent diffusion scheme Balázs Szintai (balazs.szintai@meteoswiss.ch) PBL height: good indicator of the overall performance of the turbulence scheme Following methods were applied to COSMO model outputs to diagnose the PBL height: Gradient Richardson number Bulk Richardson number Turbulent Kinetic Energy (TKE) Turbulent fluxes (momentum and heat) Theoretical approaches Unstable  Slab model (prognostic) Stable  Zilitinkevich equation (diagnostic) Methods for diagnosing PBL height Mean variables Turbulence variables

4. 4 Component testing of the COSMO model’s turbulent diffusion scheme Balázs Szintai (balazs.szintai@meteoswiss.ch) Validation – PBL heights Using radiosounding profiles (Bulk Ri method) Validation against 00 and 12 UTC soundings 4 stations (Payerne, Stuttgart, Munich, Milan) Verification periods:  10 ideal convective and stable case (in 2006-2007) Cloud free, low winds  Continuous period: 2008-02-20 – 2008-03-20 Both cyclonic and anticyclonic conditions are represented Operational setting: COSMO-7 and COSMO-2

5. 5 Component testing of the COSMO model’s turbulent diffusion scheme Balázs Szintai (balazs.szintai@meteoswiss.ch) Results – PBL heights Best results: Richardson number, Momentum flux, Theoretical approaches Problems: TKE  considerable overestimation Unstable Stable BIAS (relative) RMSE (relative)

6. 6 Component testing of the COSMO model’s turbulent diffusion scheme Balázs Szintai (balazs.szintai@meteoswiss.ch) Component testing Motivation: Better understand the behaviour of the one- equation turbulence scheme in different situations Method: component testing  analyse the TKE budget terms one by one Approach: inter-comparison of Measurements LES COSMO 3D COSMO 1D Case studies: Dry convective – Ideal case Dry convective – Real case (LITFASS)

7. 7 Component testing of the COSMO model’s turbulent diffusion scheme Balázs Szintai (balazs.szintai@meteoswiss.ch) Component testing Budget terms: I. : Local tendency II. : Advection by the mean wind III. : Buoyancy production term IV. : Shear production term V. : Turbulent transport of TKE VI. : Pressure correlation term VII. : Dissipation Prognostic equation for TKE: I. II. III. IV. V. VI. VII. Explicit handling

8. 8 Component testing of the COSMO model’s turbulent diffusion scheme Balázs Szintai (balazs.szintai@meteoswiss.ch) Ideal convective case Horizontally homogeneous and flat Dry case Constant heating rate Non-rotating Shear free Large Eddy Simulation (LES): Mironov et al., 2000 TKE budget terms are available TKE

9. 9 Component testing of the COSMO model’s turbulent diffusion scheme Balázs Szintai (balazs.szintai@meteoswiss.ch) Ideal convective case Results I. Operational COSMO-2 levels First full level at 10 m dt=72 s TKE 10 m equidistant levels First full level at 5 m dt=72 s Transport vanishes due to numerical limiter in the explicit scheme LES

10. 10 Component testing of the COSMO model’s turbulent diffusion scheme Balázs Szintai (balazs.szintai@meteoswiss.ch) Ideal convective case Results II. Solutions Keep explicit formulation Deactivate numerical limiter Smaller timestep needed 10 m levels, dt=3.6 s LES TKE Implement semi-implicit formulation for the transport term (O. Fuhrer) Results are not sensitive to timestep Same results for 20m, 10m, 1m levels

11. 11 Component testing of the COSMO model’s turbulent diffusion scheme Balázs Szintai (balazs.szintai@meteoswiss.ch) Ideal convective case - Conclusions Solution independent of vertical resolution and timestep achieved Turbulent transport of TKE is too weak Negative buoyancy flux at PBL top is missing Horizontal velocity variances are badly described at the PBL top and near the surface Implicit handling of the diffusion term in the TKE equation should be considered

12. 12 Component testing of the COSMO model’s turbulent diffusion scheme Balázs Szintai (balazs.szintai@meteoswiss.ch) LITFASS-2003 Measurement campaign conducted in the area of Lindenberg, Germany Main goal: measurement of turbulent fluxes in an inhomogeneous terrain Measurement types: Micrometeorological stations Remote sensing 100 m tower Helicopter flights 20 km

13. 13 Component testing of the COSMO model’s turbulent diffusion scheme Balázs Szintai (balazs.szintai@meteoswiss.ch) LITFASS-2003 Single column runs were made for the 8 land cover types Two types of runs: „Free run” initialized with mesurements „Forced run”: surface temperature and humidity Mean variables (wind, temp.) and surface fluxes were verified with micrometeorological stations TKE was compared to the tower measurements Humidity profile

14. 14 Component testing of the COSMO model’s turbulent diffusion scheme Balázs Szintai (balazs.szintai@meteoswiss.ch) Outlook High resolution LES runs (<10m) for the real convective case Investigation of both ideal and real stable cases